Semiconductor lasers generate light that can be used in optical communication systems, compact disc (CD) players, digital videodisc (DVD) players, scanners and other systems. Semiconductor lasers for optical communications include vertical-cavity surface-emitting lasers (VCSELs), edge-emitting lasers, in-plane emitting lasers and the like.
Until recently, relatively expensive Fabry-Perot (FP) and distributed-feedback (DFB) lasers have been used to generate light at the wavelengths presently used in the telecommunication industry for transmission via fiber-optic links. Although VCSEL technology has proven to be a viable lower-cost, lower-power alternative well suited for short-haul network applications, had difficulty to produce reliable, cost-effective VCSELs for use at the longer wavelengths used in medium- and long-haul fiber-optic communications links.
A VCSEL is composed of an active region sandwiched between vertically-stacked mirrors, commonly known as distributed Bragg reflectors (DBRs) or Bragg mirrors. The active region typically includes quantum wells that generate the light. The quantum wells are composed of thin layers of semiconductor materials that differ in band-gap energy. To achieve the necessary reflectivity, the number of semiconductor or dielectric layers constituting each of the DBRs can be quite large. The VCSEL emits the light generated in the active region through one of the mirrors, which has a reflectivity less than that of the other of the mirrors. Light is output from a VCSEL from a relatively small area on the surface of the semiconductor, directly above or below the active region.
The potential for VCSELs to generate light with relatively long wavelengths has not been realized due, in part, to the difficulty of epitaxially growing DBRs that have suitable optical, electrical, and thermal properties on an indium phosphide (InP) substrate. Two of the more significant problems are high optical losses and high joule heating in the Bragg mirror fabricated using p-type semiconductor materials.
There has been considerable interest in long wavelength VCSELs on gallium arsenide (GaAs) substrates because of the mature fabrication process for VCSELs operating at 850 nm. It has been found that adding nitrogen (N) to gallium indium arsenide (GaInAs) reduces the energy band gap enough to extend the wavelength to 1300 nm and continue to take advantage of the favorable DBR mirror technology available on GaAs substrates. However, high optical loss in p-type material is a problem.
The industry has explored incorporating a tunnel junction into a VCSEL to address these problems. Incorporating a tunnel junction allows both DBRs to be fabricated using n-type semiconductor materials. A DBR fabricated using n-type semiconductor materials has significantly lower optical losses and higher electrical conductivity than a DBR fabricated using p-type semiconductor material. Reduced optical losses lead to a lower threshold current and a correspondingly higher differential gain. Higher differential gain is an important parameter for achieving high-bandwidth modulation. High-bandwidth modulation is desirable for optical fiber-based communication systems.
FIG. 1 shows a side view of an example of a prior-art semiconductor device 100 incorporating a tunnel junction structure 112. The tunnel junction structure is composed of an n-type tunnel junction layer 104, a p-type tunnel junction layer 106 and a tunnel junction 110 between the tunnel junction layers. The n-type tunnel junction layer is a layer of an n-type semiconductor material. The p-type tunnel junction layer is a layer of a p-type semiconductor material. Applying a reverse bias across tunnel junction 110 will cause a tunneling current to flow across the tunnel junction. A reverse bias is applied by setting n-type tunnel junction layer 104 to a more positive voltage than p-type tunnel junction layer 106. It is desirable to minimize the voltage drop across the tunnel junction to reduce the overall voltage drop across the VCSEL. To minimize the voltage drop across the tunnel junction, conventional approaches have focused on maximizing the doping concentrations in the materials of the tunnel junction layers.
Also shown in FIG. 1 are n-type layer 102 on which n-type tunnel junction layer 104 is grown and p-type layer 108 grown on p-type tunnel junction layer 106; n-type layer 102 may constitute the substrate of semiconductor device 100. Alternatively, n-type layer 102 may be grown on or over the substrate.
Many different pairs of semiconductor materials that can be used as the materials of n-type tunnel junction layer 104 and of p-type tunnel junction layer 106 are known in the art. In the semiconductor device 100 illustrated in FIG. 1, the semiconductor material of n-type tunnel junction layer 104 is n-type GaAs or GaInAs and the semiconductor material of p-type tunnel junction layer 106 is GaAs or GaInAs. If GaInAs is used in either or both of the layers 104 and 106, the In concentration and/or the thickness of the tunnel junction layers should be sufficiently small so that the critical thickness is not exceeded. The critical thickness is the thickness at which crystal defects begin to form due to a mismatch in the lattice parameters of the materials. The material of layer 102 is also n-type GaAs that has a lower dopant concentration than the material of n-type tunnel junction layer 104. The material of layer 108 is also p-type GaAs that has a lower dopant concentration than the material of p-type tunnel junction layer 106.
Tunnel junctions having a low voltage drop are formed of materials that establish a large built-in electrostatic field across the tunnel junction. A large electrostatic field requires a large potential difference across a short distance, and is typically generated by using very high doping concentrations in the tunnel junction layers that minimize the width of the depletion region at the tunnel junction.
FIGS. 2A and 2B each include an energy diagram 200 and an electrical circuit model 202 that show some of the characteristics of tunnel junction structure 102. FIG. 2A shows the characteristics of the tunnel junction structure at equilibrium. FIG. 2B shows the characteristics of the tunnel junction structure under reverse bias. Each energy diagram shows the conduction band energy ECn and the valence band energy EVn of the semiconductor material of n-type tunnel junction layer 104. Each energy diagram also shows the conduction band energy ECp and the valence band energy EVp of the semiconductor material of p-type tunnel junction layer 106; n-type tunnel junction layer 104 and p-type tunnel junction layer 106 collectively form tunnel junction 110.
The energy diagram of FIG. 2A shows the depletion region 204 that exists at tunnel junction 110 at equilibrium. At equilibrium, the Fermi level EFn of the material of n-type tunnel junction layer is equal to the Fermi level EFp of the material of p-type tunnel junction layer 106. The conduction bands of the materials of the tunnel junction layers differ in energy, which establishes the built-in potential barrier 206 at the tunnel junction that prevents conduction through the tunnel junction at low forward bias. The electrostatic field strength E at the tunnel junction depends on the height of the built-in potential barrier and depends inversely on the width W of the depletion region 204 at the tunnel junction.
A forward bias applied across tunnel junction 110 decreases the height of the built-in potential barrier at the tunnel junction. Sufficient forward bias causes current to flow across the tunnel junction in the forward direction. A forward bias is established by setting p-type tunnel junction layer 106 to a more positive voltage than n-type tunnel junction layer 104. The width of depletion region 204 decreases under forward bias (not shown).
A reverse bias applied across tunnel junction 110 adds to the height of the built-in potential barrier and increases the width of depletion region 204 to W′. The reverse bias separates the Fermi levels EFn and EFp on opposite sides of the tunnel junction. In the example shown, the Fermi level EFp of the material of p-type tunnel junction layer 106 has increased relative to its equilibrium level, whereas the Fermi level EFn of the material of n-type tunnel junction layer 104 remains substantially unchanged. In a conventional p-n junction, only a small leakage current flows across the junction under reverse bias. However, in tunnel junction 110, the reverse bias causes current to flow and occurs due to electrons tunneling through the potential barrier.
The reverse bias elevates the valence band energy EVp of the material of p-type tunnel junction layer 106 above the conduction band energy ECn of the material of n-type tunnel junction layer 104. This allows electrons in the valence band of the material of the p-type tunnel junction layer to tunnel through the potential barrier to unoccupied sites in the conduction band of the material of n-type tunnel junction layer 104, as shown schematically in FIG. 2B. The greater the reverse bias applied across tunnel junction 110, the higher the probability that an electron, e−, will tunnel across tunnel junction 110, and the higher the conduction through the tunnel junction.
In a conventional tunnel junction structure 112, the semiconductor materials of tunnel junction layers 104 and 106 have a relatively large band-gap energy difference. Such materials establish a relatively high potential barrier at tunnel junction 110. The reverse bias that has to be applied across the tunnel junction structure to cause a tunneling current on the order of the laser current of a laser diode to flow is therefore relatively large. Such a tunnel junction can therefore be regarded as having a high voltage drop, which is undesirable in many applications. The voltage drop of a tunnel junction structure can be reduced by reducing the width of the depletion region to increase the tunneling probability. This approach is conventionally used to reduce the voltage drop by doping the tunnel junction layers to as high a concentration as possible.
Materials with a low band-gap may be used as the semiconductor materials of n-type tunnel junction layer 104 and p-type tunnel junction layer 106 to reduce the doping concentrations necessary to achieve a given conductivity. Low band-gap materials, such as GaInAs, Eg=0.75 eV), have been used in InP-based devices, i.e., semiconductor devices with substrates of indium phosphide, to achieve a relatively high tunnel-junction conductivity. Low bandgap materials are also available in GaAs-based devices. However, doping concentrations sufficient to ensure a high probability of tunneling are difficult to achieve in these materials.
The addition of N to the n-type tunnel junction layer lowers the bandgap of the tunnel junction layer and improves the tunneling probability.
Prior U.S. patent application Ser. No. 10/243,520 discloses a tunnel junction structure in which a fraction of the arsenic atoms in the semiconductor material of the p-type tunnel junction layer are replaced with antimony atoms, thus increasing the valence band energy of the p-type material. The tunnel junction structure of prior application Ser. No. 10/243,520 also discloses replacing a fraction of the arsenic atoms in the semiconductor material of the n-type tunnel junction layer with nitrogen atoms, thus decreasing the conduction band energy of the n-type material. Increasing the valence band energy of the semiconductor material of the p-type tunnel junction layer and decreasing the conduction band energy of the semiconductor material of the n-type tunnel junction layer increase the probability of tunneling occurring at a given reverse bias applied across the tunnel junction. Unfortunately, when nitrogen is added to the material of the n-type tunnel junction layer, the reduction in the voltage drop across the tunnel junction is less than predicted.
Thus, what is needed is a tunnel junction structure that can be fabricated using the GaAs material system in which the tunneling probability is increased and, hence, the conductivity of the tunnel junction structure is increased and the voltage drop across the tunnel junction is reduced.